Bryan Williams: Iraq War
Bryan Williams has been doing independent analyses of various events based on the 15-minute blocked data. He usually looks at the same period as is specified in the formal prediction, and sometimes examines longer periods around that time for context. In this page, I assemble his analysis of the peace demonstrations on and then a collection of analyses of the beginning and the end of the Iraq war. I include his emailed description of his work.
Peace Demonstrations Feb 15 2003
Here is the 15-minute block result for the global peace demonstrations on February 15th. It is interesting that towards the end the trend takes on a positive direction that is similar to the formal seconds result. The result was found to have Chi-square = 4762.38, 4656 df, p = .135.
The Beginning and the End of the War in Iraq
Attached are the graphical results for the events of the Iraq war in 15-minute resolution, as I had promised.
The first graph is one that I made using the preliminary prediction for the start of the war that had initially been on the formal specification registry around March 19. I was watching the news coverage of the events that night in order to get exact times for them, and NBC News had reported that the air raid sirens had started sounding in Baghdad (which was assumed to be the beginning of the bombing by U.S. and British warplanes) at around 9:30 PM EST (2:30 UTC), which is what is marked on the expectation line. Since this was a makeshift prediction, I didn't calculate a Chi-Square value for it, and create it just for a first look at the data.
The second graph shows the formal prediction period in 15-minutes, which has a trend quite similar to the seconds result. The first tickmark is again the reported time of air raid sirens sounding, the second mark is the time of President Bush's address to the nation, and the third is the second wave of the Baghdad bombing. The results were a Chi-Square of 1495.46, 1507 df, p = .578.
The third graph shows the first full day of the war in 15-minute resolution, with the marked events on the expectation line being the same as those on the seconds graph. From a visual standpoint I thought the trend was very impressive, and if it were a formal prediction, it would have had Chi-Square = 4632.7, 4483 df, p = .058.
The fourth graph is for the formal prediction made for the end of the war and the fall of the Saddam Hussein statue on April 9. From the visual standpoint, I thought it was interesting how trend surrounding the time of the Saddam statue falling seemed to be in line with Norm Hirsh's idea of a "blip" in the data. The hour prediction resulted in Chi-Square = 188.68, 196 df, p = .633.
The last graph is the one I did the day following the fall of Baghdad on April 9, which was a preliminary context exploration using a makeshift prediction for the full UTC day. The first tickmark is the time reported by NBC News as being the approximate moment U.S. troops entered Baghdad, and the second marks the fall of the Saddam statue, which I had to approximate at that time because I wasn't able to watch the full coverage.
Sorry that these are coming so late; I was busy late in the semester and was traveling during May.
Bryan Williams, UNM
The End of the War; Saddam Falls